The nodes form a test dataset prior to the formation of a new block cycle. After the formation of the test sample, the matrices are checked. The matrix with the smallest error wins and gets loaded into the next era.The Velas blockchain platform uses neural networks that are optimized by artificial intuition to improve consensus algorithms. Velas's goal is to overcome and improve existing problems and challenges faced by most blockchain. The neural network is used to calculate the rewards from the node operator and the time of block formation. The calculation server matrix (neural network weights) is at the node of the network member to receive prizes. Similar to Bitcoin miners, nodes must be large computing resources to calculate matrices. For example, installing a special powerful video card (GPU). To train neural networks, genetic algorithms are used during the pre-training period and use the error propagation method to find the minimum objective function. The node of the winning matrix receives Velas (VLX) from the network by solving blockchain network maintenance tasks, each AIDPOS node makes an intellectual contribution by further developing a neural network. This all happens without requiring any understanding programming languages by operators of selected nodes.
the node forms a test dataset before the formation of a new block cycle. After forming the test sample, the matrix is examined. The matrix with the smallest error wins and is loaded into the next era. The winning node's node receives Velas (VLX) from the network by completing the task of maintaining the blockchain network, each AIDPOS node makes an intellectual contribution by further developing the neural network. This all happens without requiring an understanding of the programming language by the operator of the selected node. Network Training Algorithm Each node forms its own dataset (training sample) from the blockchain data. In this data, layered training of genetic algorithms occurs. Each subsequent layer studies the previous one: training on the autoencoder principle. After pre-training, a backpropagation method is applied to tighten the weights to a minimum. The network must be trained before the next block cycle.
Genetic algorithm:
Creation of a specimen: a matrix with random weights (genes)
Competition: getting the minimum of the objective function
Selection: ranking specimens by error. The smallest error is a victory.
Reproduction: exchange of the matrix’ elements, or genes, 50/50 from the first two most successful.
The cycle is repeated until a 70% probability is obtained
Formation of an example: a grid with arbitrary loads (genes)Competition: getting the base of the goal functionSelection: positioning examples by mistake. The littlest blunder is a victory.Reproduction: trade of the framework' components, or qualities, 50/50 from the initial two most successful.The cycle is rehashed until a 70% likelihood is gotten. Picking the best system The hubs structure a test dataset preceding the development of another square cycle.After the arrangement of the test, the frameworks are checked.The lattice with the littlest blunder wins and gets stacked into the following era.The hub of the triumphant grid gets Velas (VLX) from the system. A business opportunity for the deal and acquisition of the best examples will advance which will open up the improvement of the blockchain arrange upkeep calculation with remunerations for different frameworks dependent on Velas.
Website: https://velas.com/
White Paper: https://velas.com/VELAS-Technical_Paper.pdf
Twitter: https://www.twitter.com/VelasBlockchain/
Telegram: https://t.me/velascommunity
Facebook: https://www.facebook.com/velasblockchain/
Instagram: https://www.instagram.com/velas.blockchain/
Medium: https://medium.com/@VelasBlockchain
Linkedin: https://www.linkedin.com/company/velas-ag
Github: https://github.com/velas
Discord: https://discord.gg/CTcKpPc
Bitcointalk username: farukahmed
Bitcointalk url: https://bitcointalk.org/index.php?act...
Telegram username: @farukahmed
VLX wallet address: VLMQzV1sysRRR9eHcAMBYAFYMUBTN1HKPTq